We propose an image segmentation algorithm to solve the problem of fast image segmentation when surface unmanned vehicles perform target tracking and recognition tasks. Firstly, the algorithm uses the mean filtering algorithm to filter the RGB ocean background image, and uses its nonparametric property to obtain the clustering center of the image and the number of categories, which are used as the initialization parameters to perform fuzzy C-means clustering on the image.  On this basis, the image is binarized by the Otsu method to realize target extraction. The BSDS500 standard dataset and marine background images are used to verify itse segmentation effect and efficiency. Comparison with the traditional fuzzy C-means algorithm, pulse coupled neural network algorithm, adaptive genetic algorithm and Markov random field algorithm, proves the effectiveness of the proposed algorithm.